Speed Over Spectacle: A16z Report Reveals How Startups Are Really Using AI

Startups Fuel AI Adoption by Chasing Speed, Not Hype

SAN FRANCISCO, CA - OCTOBER 9, 2025 - Startups are not just integrating artificial intelligence; they are swiftly gaining speed and efficiency through AI tools, deeply reshaping operational models, according to a recent analysis by venture capital firm Andreessen Horowitz (a16z) and fintech company Mercury. The combined study, which examined AI-related transactions from more than 200,000 startups between June and August 2025, uncovers a clear move away from experimental AI adoption toward targeted spending on tools that tangibly speed up workflows across varied functions, from note-taking and design to coding and communication.

This insight, outlined in a16z's "The AI Application Spending Report," provides a real-time indicator of what early-stage firms are actually buying in the AI arena. In contrast to spending on core AI infrastructure-compute, models, and developer utilities-these outlays highlight the hands-on use of AI within product development and everyday operations. The results are consistent with wider surveys showing that early-stage firms not only intend to boost their AI budgets but also see a greater return on investment (ROI) versus conventional tools.

The Dominance of "For Everyone" AI Tools

A large share, approximately 60%, of AI expenditure by these startups is directed towards "for everyone" products. This category includes prominent names such as OpenAI, Anthropic, Perplexity, Notion, and Manus. These solutions, frequently described as assistants or "smart workspaces," form a fluid and leader-less competitive arena, signalling broad demand for general-purpose AI augmentation instead of narrowly-focused offerings.

Specific horizontal applications attracting significant spend include:

• OpenAI (ranked #1 in spend volume)
• Anthropic (ranked #2)
• Perplexity (ranked #12)
• Notion (ranked #10)
• Manus (ranked #33)
• Merlin AI (ranked #30)
• Fyxer (ranked #7), known for email drafting.
• Happyscribe (ranked #36)
• Plaude (ranked #38), a manufacturer of a hardware pin
• Otter AI (ranked #41)
• Read AI (ranked #49)
• Cluely (ranked #26)

The Creative Stack: Content Creation as a Core Function

The analysis highlights that the creative stack is no longer a peripheral concern but a core element of everyday operations. Tools like Freepik, ElevenLabs, Canva, Photoroom, Midjourney, Descript, Opus Clip, and CapCut are no longer niche applications but essential components that enable rapid content generation. This suggests that content creation is increasingly woven into a variety of roles rather than staying the exclusive domain of dedicated creative teams. Freepik (#4) leads this category, with ElevenLabs (#5) following closely, particularly in audio. Image and video generation tools also feature prominently, with Midjourney (#28) still generating the majority of consumer revenue while showing notable enterprise adoption. New entrants such as Arcads (#47) and Tavus (#50) are likewise emerging for avatar creation, especially for advertising and multi-purpose uses.

"Vibe-Coding" Tools Accelerate Development Cycles

In software development, "vibe-coding" tools like Replit, Cursor, Lovable, and Emergent have entered the mainstream, indicating a clear tilt toward accelerated prototyping over traditional large development squads. Replit (#3) stands out as the top product in this segment, producing roughly 15 times more revenue from Mercury customers than Lovable, which recorded higher consumer web traffic but lower enterprise spend. This gap illustrates startups' readiness to invest in developer tools that directly translate into faster product cycles. Replit, identified as an agentic product development tool, exemplifies this pattern.

Vertical AI: The Rise of "AI Employees"

Vertical AI applications are evolving into sophisticated "AI employees," a trend that points to a strategic shift toward automation rather than conventional hiring or contracting. Of the seventeen vertical-application companies identified, five are explicitly built to execute workflows end-to-end. These include:

• Crosby Legal (#27), functioning as an agentic law firm.
• Cognition (#34), focusing on AI engineering tasks.
• 11x (#37), providing automated go-to-market (GTM) employees.
• Serval (#39), an AI IT service desk.
• Alma (#42), offering AI-powered immigration law services.

Beyond these "AI employee" solutions, vertical AI also backs human augmentation in areas such as customer service (Lorikeet #8, Customer.io #14, Ada #40, Crisp #46), sales/GTM (Instantly #13, Clay #25, 11x #37), recruiting/HR (Micro1 #9, Metaview #19, Applaud #43), and operations (Delve #11 for compliance automation, Combinely #29 for accounting).

The B2C-to-B2B Fast Track

A striking pattern observed is the rapid migration of consumer-focused AI applications into enterprise offerings. Nearly 70% of the top AI tools identified began as consumer apps, scaling their B2C-to-B2B trajectory within a compressed window of one to two years. This swift path underscores the agility and inherent scalability of AI-native products, as well as the immediate value they deliver-often first validated by individual users before being embraced by enterprises. Notable examples include Cluely and Midjourney, which continue to pull substantial consumer revenue even as they penetrate the enterprise market. OpenAI itself, as of last October, saw 75% of its revenue from consumers; that figure has recently shifted to an estimated 50/50 split, further illustrating this evolution.

The data, gathered via Mercury's financial transaction analysis (excluding non-Mercury card spend and Mercury Personal customers), offers a robust, spend-centric perspective that contrasts with mere web-traffic reports. It definitively shows that for startups, AI investment is now measured solely by its ability to deliver rapid, tangible outcomes and drive efficiency.